Research Team: Alexander Skabardonis (lead) and Qijian Gan
UC Campus(es): UC Berkeley
Problem Statement: Arterial traffic state estimation becomes a difficult task in traffic simulation as the network size gets bigger. Even with a long warm-up period, it is not guaranteed for the simulation tool to drive the initial traffic states to the correct ones. A more direct approach is required to estimate the traffic states from the field data and place the right number of vehicles at intersection approaches before running simulation. For evaluation of intersection performance, metrics of delay and Level of Service (LOS) are often used. However, estimates computed from the state-of-the-practice Highway Capacity Manual (HCM) method are not reliable under heavy traffic conditions. Therefore, instead of using indirect metrics like delay and LOS, it is more important to come up with a method to directly estimate the traffic states so as to evaluate the intersection performance properly.
Project Description: This project developed a novel approach to estimating the traffic states on arterial road links controlled by signalized intersections using both loop detector data and signal phasing information. A trapezoidal fundamental diagram was derived that includes two occupancy thresholds to categorize the traffic states into three different regimes: uncongested, congested, and downstream queue spillback. The parameters used to compute these two thresholds are closely related to road geometry, detector layout, signal settings, and vehicle dynamics, which can be obtained from the field data. A case study was performed using field data from three intersections along Huntington Drive in the City of Arcadia in the I-210 Connected Corridors pilot. Flow-occupancy relations were obtained for both advance and stopline detectors at the intersection approaches, and it was found that advance detector information is more reliable because it is less impacted by the traffic signal. The proposed trapezoidal fundamental diagram was validated using a dataset with six months of detector data.
Project Partner(s): Caltrans, Los Angeles County Metropolitan Transportation Authority